AP Calculus BC

Improper Integrals: Convergence and Divergence

An improper integral has an infinite limit of integration or an integrand that is unbounded somewhere on the interval. Replace the trouble spot with a variable, integrate, then take a limit. If the limit is a finite number, the integral converges to that number. Otherwise it diverges.

Two ways an integral can be improper

The Fundamental Theorem of Calculus evaluates ∫abf(x) dx\int_a^b f(x)\,dx when [a,b][a, b] is a finite interval and ff is continuous on it. An integral is improper when the interval is infinite or the integrand is unbounded somewhere on it. Topic 6.13, Evaluating Improper Integrals, is a BC-only topic.

TypeWhat goes wrongExample
Infinite limit of integrationThe interval has infinite length∫1∞1x3 dx\int_1^{\infty} \frac{1}{x^3}\,dx
Unbounded integrandThe integrand has a vertical asymptote at an endpoint or inside the interval∫041x dx\int_0^4 \frac{1}{\sqrt{x}}\,dx

In both cases the fix is the same: move the problem into a limit. An improper integral converges when that limit exists as a finite number, and the number is the value of the integral. It diverges when the limit is infinite or fails to exist.

Infinite limits of integration

Replace the infinite limit with a variable, integrate over the finite interval, and let the variable go to infinity:

∫a∞f(x) dx=lim⁡b→∞∫abf(x) dx∫−∞bf(x) dx=lim⁡a→−∞∫abf(x) dx\int_a^{\infty} f(x)\,dx = \lim_{b \to \infty} \int_a^b f(x)\,dx \qquad \int_{-\infty}^{b} f(x)\,dx = \lim_{a \to -\infty} \int_a^b f(x)\,dx

When both limits are infinite, split the integral at any convenient number cc and treat each half as its own limit. The whole integral converges only if both halves converge:

∫−∞∞f(x) dx=∫−∞cf(x) dx+∫c∞f(x) dx\int_{-\infty}^{\infty} f(x)\,dx = \int_{-\infty}^{c} f(x)\,dx + \int_{c}^{\infty} f(x)\,dx

A divergent example shows why the limit matters. For ∫1∞1x dx\int_1^{\infty} \frac{1}{x}\,dx, the finite piece is ∫1b1x dx=ln⁡b\int_1^b \frac{1}{x}\,dx = \ln b, and ln⁡b\ln b grows without bound as b→∞b \to \infty. The integral diverges even though the integrand 1x\frac{1}{x} approaches 00.

Unbounded integrands

If the integrand has a vertical asymptote at an endpoint, approach that endpoint from inside the interval with a one-sided limit. When the asymptote is at the left endpoint aa:

∫abf(x) dx=lim⁡t→a+∫tbf(x) dx\int_a^b f(x)\,dx = \lim_{t \to a^+} \int_t^b f(x)\,dx

When the asymptote is at the right endpoint bb, use lim⁡t→b−∫atf(x) dx\lim_{t \to b^-} \int_a^t f(x)\,dx. When the asymptote is at a point cc inside the interval, split the integral at cc and handle the two pieces separately. Both pieces must converge for the whole integral to converge.

The hidden asymptote

The integral ∫−111x2 dx\int_{-1}^{1} \frac{1}{x^2}\,dx looks ordinary, and plugging the bounds into −1x-\frac{1}{x} gives −1−1=−2-1 - 1 = -2. That answer is wrong: a positive integrand cannot have a negative integral. The integrand is unbounded at x=0x = 0, inside the interval. Splitting at 00 gives lim⁡t→0+∫t11x2 dx=lim⁡t→0+(1t−1)=∞\lim_{t \to 0^+} \int_t^1 \frac{1}{x^2}\,dx = \lim_{t \to 0^+} \left(\frac{1}{t} - 1\right) = \infty, so the integral diverges.

Before evaluating any definite integral of a quotient, a root, or a logarithm, check whether the integrand is unbounded anywhere on the interval, including the endpoints.

The benchmark integrals

Powers of xx give two families worth knowing, because many other integrals behave like one of them. The cutoff in both families is p=1p = 1, and the two families behave in opposite ways.

IntegralConverges whenValue when it convergesDiverges when
∫1∞1xp dx\int_1^{\infty} \frac{1}{x^p}\,dxp>1p > 11p−1\frac{1}{p - 1}p≤1p \le 1
∫011xp dx\int_0^1 \frac{1}{x^p}\,dxp<1p < 111−p\frac{1}{1 - p}p≥1p \ge 1

The function 1x\frac{1}{x}, with p=1p = 1, diverges in both directions. Out toward infinity, a larger power makes the integrand shrink faster, which helps convergence. Near 00, a larger power makes the integrand blow up faster, which hurts it.

Exponential decay also converges. For any k>0k > 0, ∫0∞e−kx dx=lim⁡b→∞(1k−e−kbk)=1k\int_0^{\infty} e^{-kx}\,dx = \lim_{b \to \infty} \left(\frac{1}{k} - \frac{e^{-kb}}{k}\right) = \frac{1}{k}. The first family reappears in Unit 10, where the same cutoff decides which pp-series converge. See improper integral vs infinite series for that connection.

Common mistakes

  • Treating ∞\infty as a number. Writing [−1x]1∞\left[-\frac{1}{x}\right]_1^{\infty} and plugging in infinity skips the limit. Write the limit explicitly and carry it until the last step.
  • Missing an asymptote inside the interval and applying the Fundamental Theorem across it.
  • Concluding that an integral converges because the integrand goes to 00. The integrand 1x\frac{1}{x} goes to 00 and its integral on [1,∞)[1, \infty) still diverges.
  • Letting two divergent pieces cancel. If either piece of a split integral diverges, the whole integral diverges.
  • Using a two-sided limit at an endpoint. The variable must approach from inside the interval.

Quick check

If the limit at the end needs a form like ∞∞\frac{\infty}{\infty} resolved, as with lim⁡b→∞beb\lim_{b \to \infty} \frac{b}{e^b}, use L'Hopital's Rule on that limit and then finish the evaluation.

Worked examples

Worked example

An infinite limit of integration

Evaluate ∫1∞1x3 dx\int_1^{\infty} \frac{1}{x^3}\,dx or show that it diverges.

  1. Rewrite as a limit: ∫1∞1x3 dx=lim⁡b→∞∫1bx−3 dx\int_1^{\infty} \frac{1}{x^3}\,dx = \lim_{b \to \infty} \int_1^b x^{-3}\,dx.
  2. Integrate on the finite interval: ∫1bx−3 dx=[−12x2]1b=−12b2+12\int_1^b x^{-3}\,dx = \left[-\frac{1}{2x^2}\right]_1^b = -\frac{1}{2b^2} + \frac{1}{2}.
  3. Take the limit: as b→∞b \to \infty, 12b2→0\frac{1}{2b^2} \to 0, so the limit is 12\frac{1}{2}.
  4. Compare with the benchmark: p=3>1p = 3 > 1, and 1p−1=12\frac{1}{p - 1} = \frac{1}{2}, which agrees.

The integral converges to 12\frac{1}{2}.

Worked example

An integrand unbounded at an endpoint

Evaluate ∫041x dx\int_0^4 \frac{1}{\sqrt{x}}\,dx or show that it diverges.

  1. Find the trouble spot. The integrand 1x\frac{1}{\sqrt{x}} is unbounded as x→0+x \to 0^+, and 00 is the left endpoint.
  2. Rewrite as a one-sided limit: ∫04x−1/2 dx=lim⁡t→0+∫t4x−1/2 dx\int_0^4 x^{-1/2}\,dx = \lim_{t \to 0^+} \int_t^4 x^{-1/2}\,dx.
  3. Integrate: ∫t4x−1/2 dx=[2x]t4=4−2t\int_t^4 x^{-1/2}\,dx = \left[2\sqrt{x}\right]_t^4 = 4 - 2\sqrt{t}.
  4. Take the limit: as t→0+t \to 0^+, 2t→02\sqrt{t} \to 0, so the limit is 44.

The integral converges to 44.

Worked example

An asymptote inside the interval

Evaluate ∫031(x−1)2/3 dx\int_0^3 \frac{1}{(x - 1)^{2/3}}\,dx or show that it diverges.

  1. Find the trouble spot. The integrand is unbounded at x=1x = 1, which is inside [0,3][0, 3]. Split there: ∫011(x−1)2/3 dx+∫131(x−1)2/3 dx\int_0^1 \frac{1}{(x - 1)^{2/3}}\,dx + \int_1^3 \frac{1}{(x - 1)^{2/3}}\,dx.
  2. Find an antiderivative: ∫(x−1)−2/3 dx=3(x−1)1/3+C\int (x - 1)^{-2/3}\,dx = 3(x - 1)^{1/3} + C.
  3. Left piece: lim⁡t→1−[3(x−1)1/3]0t=lim⁡t→1−(3(t−1)1/3−3(−1)1/3)=0−3(−1)=3\lim_{t \to 1^-} \left[3(x - 1)^{1/3}\right]_0^t = \lim_{t \to 1^-} \left(3(t - 1)^{1/3} - 3(-1)^{1/3}\right) = 0 - 3(-1) = 3.
  4. Right piece: lim⁡t→1+[3(x−1)1/3]t3=lim⁡t→1+(3⋅21/3−3(t−1)1/3)=323\lim_{t \to 1^+} \left[3(x - 1)^{1/3}\right]_t^3 = \lim_{t \to 1^+} \left(3 \cdot 2^{1/3} - 3(t - 1)^{1/3}\right) = 3\sqrt[3]{2}.
  5. Both pieces converge, so add them: 3+323≈6.7803 + 3\sqrt[3]{2} \approx 6.780.

The integral converges to 3+323≈6.7803 + 3\sqrt[3]{2} \approx 6.780.

Frequently asked questions

If the integrand goes to zero, does the improper integral converge?

Not necessarily. The integrand has to shrink fast enough. 1x\frac{1}{x} approaches 00 as x→∞x \to \infty, yet ∫1∞1x dx\int_1^{\infty} \frac{1}{x}\,dx diverges because ln⁡b\ln b grows without bound. 1x2\frac{1}{x^2} shrinks faster and its integral on [1,∞)[1, \infty) converges to 11.

How do I tell that an integral is improper when the limits are finite?

Check whether the integrand is unbounded anywhere on the interval, endpoints included. Look for values of xx that make a denominator zero or that sit at the edge of a logarithm's domain. If there is a vertical asymptote on the interval, the integral is improper and needs a limit there.

Can I evaluate an integral from negative infinity to infinity with one symmetric limit?

No. Split it at a point and take the two limits separately. For example, ∫−bbx dx=0\int_{-b}^{b} x\,dx = 0 for every bb, but ∫0∞x dx\int_0^{\infty} x\,dx diverges, so ∫−∞∞x dx\int_{-\infty}^{\infty} x\,dx diverges.

Where this shows up outside calculus class